Artificial intelligence found 56 new attractive lens candidates

Astronomers in Europe are using artificial intelligence (AI) to find many new gravitational lens candidates in the universe. This is also the technology that Tesla is using to develop a self-driving car system.

A team of astronomers from the universities of Groningen, Naples and Bonn developed a new method to find attractive lenses .

When a galaxy is hidden behind another galaxy (when viewed from Earth), we will see the galaxy behind is a circle around the front galaxy. This phenomenon is called gravitational lens, because it is based on the theory of relativity that Einstein found, that mass can bend light .

This is a special gravitational lens, called the Einstein Ring.In this image, the gravitational force of the red glowing galaxy distorts light from the green galaxy, creating a nearly circular shape.The image was taken in 2011 by the Hubble Space Telescope, the object identified as LRG 3-757 also known as the Cosmic Horseshoe.(Image: NASA / APOD).

Artificial intelligence like big technology companies

Astronomers are always looking for gravitational lenses because it can help further study dark matter. Hunting for attractive lenses is done very often, but it is not easy to find it. Scientists have to search for thousands of images of deep space to find attractive lenses.

Until now, astronomers had to ask the help of volunteers around the world. This manual search is still heavily dependent on the amount of new images taken, ie if there is no new image taken, there will be nothing for volunteers to search for.

But the problem is not that there is no new image, but the development of telescopes too fast, observatories across the sky and create millions of new images. People have limited capacity, we cannot strain our eyes to find in every image of that huge pile of data.

Graphics show that gravitational lenses are a phenomenon described by Einstein's theory of relativity, which occurs by the bending of light.The gravitational field of a high-mass object will distort and radiate light rays away, while light rays passing through the gravitational field closer will be bent and gathered to one point.If Earth is the focus of light rays, we will see an attractive lens.(Graphics: NASA / ESA).

To solve the problem of increasing number of images, astronomers used " convolutional neural networks" . Google has also used such artificial neural networks to win a world-class go-go grandmaster. Facebook also does the same to identify what content users post with content. And Tesla is developing its self-driving car system with this method.

Astronomers have taught artificial neural networks by loading millions of available gravitational lens images, letting them practice their work by finding the same thing in millions of photos of the sky. Despite being millions of images, it is just a small sky, about 255 square feet - less than half the sky.

This is an image astronomers use to train artificial intelligence how to identify attractive lenses in practice.(Image: Enrico Petrillo / Rijksuniversiteit Groningen / Astronomie.nl).

Attractive lens candidates

Initially, this artificial nervous system found 761 attractive lens candidates. After being visually inspected by astronomers themselves, they filtered out and picked up exactly 56 attractive lenses again. However, they are still candidates because they need to have observations of large telescopes like the Hubble Space Telescope to confirm them.

Attractive lenses are found in reality by artificial intelligence.(Image: Enrico Petrillo / Rijksuniversiteit Groningen / Astronomie.nl).

In addition, this artificial intelligence system rediscovered two previously known gravitational lenses, but missed out on an outstandingly well-known gravitational lens. This overlooked gravitational lens is a small and rather complex size, the artificial neural network has not been learned about it, so it has been ignored.

In the future, researchers will further train the system to detect more small and complex gravitational lenses, as well as identify and eliminate similar objects but not. The ultimate goal that scientists aim for is that computers do it on their own without having to be checked by humans.

Surveying the thousands of degrees of sky

Scientist Carlo Enrico Petrillo from the University of Groningen in the Netherlands, author of the study, said: 'This is the first time an artificial intelligence system has been used for astronomical surveys. I think it will soon become a new benchmark for future sky surveys. Technology will survey and produce large amounts of data because we don't have enough astronomers to do all these things. "

This research project is named a project of Surveying the thousands of degrees of sky . The project uses the Very Large Telescope (VLT) of the Southern European Observatory (ESO) in Paranal Mountain in Chile, with panoramic camera OmegaCAM used to scan the sky developed by the Dutch government.

This study resulted in 56 attractive lens candidates, published in the November issue of the Royal Astronomical Society of England.